Skip to content

Commit 8f9a42d

Browse files
committed
docs(spec): anchor proof 4's corrected counts to the tree they were read on
Both corrected figures now say WHICH tree they are a reading of, in the same sentence as the number. A count anchored to a commit stays true when the graph moves; an unanchored one silently becomes a claim about a tree nobody measured, which is the same failure mode one level up from the one being fixed here. The not-delivered partition in particular is a fact about the graph, not a property of the proof — closing an open door moves it — so the docblock now tells its next reader to re-measure rather than re-date these counts. Claude-Session: https://claude.ai/code/session_01JbZnqu8bt6YqfJsr9vaFb3 Co-authored-by: Claude <noreply@anthropic.com>
1 parent 45fa837 commit 8f9a42d

1 file changed

Lines changed: 16 additions & 13 deletions

File tree

‎packages/spec/scripts/build-schemas.ts‎

Lines changed: 16 additions & 13 deletions
Original file line numberDiff line numberDiff line change
@@ -1558,20 +1558,23 @@ interface GuidanceRoutes {
15581558
* key in it, a union or pipe the probe cannot drive to a single door, or a refusal
15591559
* whose message does not carry the declared text — all read as "no evidence",
15601560
* never as "proved". ⛔ And the verdict for the second half says only THAT the
1561-
* prescription did not arrive, never WHY: that state holds 9 keys on 4 defs, and
1562-
* 3 of those 4 defs are unions carrying 7 of the 9 — for a union "the door is
1563-
* open" would be a guess this gate has not measured, the mistake this proof's
1564-
* first cut made about `additionalProperties` and must not repeat one layer down.
1561+
* prescription did not arrive, never WHY: at `88aa326deb` that state held 9 keys
1562+
* on 4 defs, 3 of those 4 defs unions carrying 7 of the 9 keys — for a union
1563+
* "the door is open" would be a guess this gate has not measured, the mistake
1564+
* this proof's first cut made about `additionalProperties` and must not repeat
1565+
* one layer down.
15651566
*
1566-
* Every count above was re-measured with THIS instrument on
1567-
* objectstack-ai/objectstack `88aa326deb` (#18579), and each is written beside
1568-
* the population it counts, because that is the one thing the first census run
1569-
* did not do: of the 1527 emitted defs, 258 resolve to exactly one declaration,
1570-
* and 147 of those 258 name a key the def does not declare; those 147 defs
1571-
* promise 779 keys, of which 770 are delivered and the 9 above are not. ⛔ Do
1572-
* not re-derive any of these from the #18301 PR body: its census row gives the
1573-
* 147 population the 258 count, and a count with no population is the defect this
1574-
* docblock was corrected for, not a figure to carry forward.
1567+
* Every count above is a reading of ONE tree — objectstack-ai/objectstack
1568+
* `88aa326deb`, where #18579 re-took all of them with THIS instrument — and each
1569+
* is written beside the population it counts: of the 1527 emitted defs, 258
1570+
* resolve to exactly one declaration, and 147 of those 258 name a key the def
1571+
* does not declare; those 147 defs promise 779 keys, of which 770 are delivered
1572+
* and the 9 above are not. Which defs sit in that last group is a fact about the
1573+
* graph at that commit and not a property of this proof — closing an open door
1574+
* moves it — so a later reader RE-MEASURES rather than re-dates these, and ⛔
1575+
* derives none of them from the #18301 PR body, whose census row gives the 147
1576+
* population the 258 count: a count wearing another population's label is the
1577+
* defect this docblock was corrected for.
15751578
* `integration/DataSyncConfig` has NO route (its shape is not a `strictObject`
15761579
* and nothing prescribes for `schedule`), so this proof does not reach the
15771580
* 2026-09-10 ruling that withheld that tombstone.

0 commit comments

Comments
 (0)